Skip to content
Fast Company Executive Board

The Fast Company Executive Board is a private, fee-based network of influential leaders, experts, executives, and entrepreneurs who share their insights with our audience.

How to fix your AI search problem

Leaders must review the company website as an AI does.

[Photo: Getty Images]

Profile photo of Kevin C. Roy

I’ve worked in SEO for 15 years, and for most of that time, the formula was straightforward. We built a strong page around topical relevance, earned authority, ranked in Google, and won the click.

That formula still matters, but it’s no longer the whole game.

AI search does not always send the user to your website. Instead, it finds pieces of evidence across the web, decides what it trusts, and builds the answer itself. A company can have the right information online, rank in the top three results, and still be left out because the LLM never found the company—or couldn’t confidently connect the claim to the company.

I see this constantly when reviewing websites. Outdated blog posts with irrelevant internal links, service pages buried in vague marketing language, unlabeled infographics, stale executive bios, and case studies omitting necessary details for results to be credible.

The company knows what all of this means, but the retrieval system has to guess, and that’s the problem. The stakes are higher today as the commercial incentives grow. Adobe reported that AI-driven traffic to U.S. retail sites increased 393% year over year during the first quarter of 2026. In March, those visitors converted 42% better than non-AI traffic.

How can you land among those impressive statistics? The first reaction to generative AI was to publish more content. I think that is exactly backward when retrieval is the bottleneck.

Why AI doesn’t see your site

Semrush found that almost 62% of the AI citations in one study were “ghost citations.” As in, the platform linked to a company’s content but did not name the brand in the generated answer. Think about how strange that is. Your information is good enough to use in an answer, but not clear enough for your identity to survive the answer.

The AI system still has to locate the answer on your site, understand which entity it belongs to, determine its currency, and decide whether it is strong enough to cite. That is a much higher bar than “the information exists somewhere on our site.”

More disconnected pages create more conflicting paths, outdated claims, and opportunities for the wrong source to be retrieved.

AI retrieval systems don’t enter through your homepage, understand the navigation, click through several pages, and connect the dots. Google explains that AI Overviews and AI Mode can use “query fan-out,” issuing multiple related searches across subtopics and data sources before assembling a response. Your web page becomes part of a collection of potential evidence, which changes what a well-built website looks like.

Picture a case-study result separated from its context: a 200% lift in qualified leads, tucked into an undated page with no client category, baseline, or explanation of the work. The number may be real, but to an answer engine it is just a fragment—hard to attribute, assess, or safely cite.

Evidence architecture keeps the claim, the proof, and the company’s identity together, so a human or machine can see what happened, who made it happen, why the result is credible, and whether it is still current.

Build an AI-friendlier search site

You can’t trick AI search by involving more keywords, hundreds of AI-written articles, or some magic schema package. Google says no special schema is required for AI Overviews or AI Mode.

Instead, you should stop making the reader—or the machine—do unnecessary detective work. Any site claim should survive outside its page. Here is a common case-study sentence:

“We increased qualified leads by 200%.”

While this sounds impressive, it’s also missing almost everything an answer engine needs. Who is “we”? What company produced the result? What kind of client was involved? What counted as a qualified lead? Over what period? Compared with what baseline?

A stronger statement that reduces ambiguity carries its own context. It identifies the business, client category, work performed, metric, timeframe, and meaningful limitations.

For example:

“An SEO firm helped a U.S.-based warehouse-automation software company increase sales-qualified organic leads by 200% from January through June 2026, compared with the preceding six months, after reorganizing its product pages and publishing customer-specific case studies. A lead was counted only after the client’s sales team accepted it. This result reflects one client engagement and does not guarantee comparable results for other companies.”

The evidence and the brand must always travel together, including author names, company names, credentials, dates, methodology, and entity relationships.

Apply the same principle around specifics to:

  • Executive biographies
  • Product specifications
  • Service areas
  • Research findings
  • Pricing qualifications
  • Awards
  • Statistics

Any important claims should sit close to their evidence and clearly connect them to the right company, person, product, or location. You want to create a mountain of information with clear primary pages, useful supporting content, deliberate internal links, current facts, named experts, and third-party proof reinforcing the claims that matter.

The business case for AI search is to become the source the answer engine can retrieve without confusion, understand without guessing, and cite without losing your name.

Kevin C. Roy is the CEO of GreenBanana SEO.